Hidden Markov Models for Vehicle Tracking with Bluetooth

Bluetooth is a short range communication protocol. Bluetooth-enabled devices can be detected using road-side equipment, and each detected device reports a unique identifier. These unique identifiers can be used to track vehicles through road networks over time. The focus of this paper is on reconstructing the paths of vehicles through a road network using Bluetooth detection data. A method is proposed that uses Hidden Markov Models, which are a well-known tool for statistical pattern recognition. The proposed method is evaluated on a mixture of real and synthetic Bluetooth data with GPS ground truth, and it outperforms a simple deterministic strategy by a large margin (30%-50%) in this case.